CLUSTERATLAS
A field guide to supercomputing

Alps

Swiss National Supercomputing Centre

CPU Accelerator Interconnect
i
Reading this map

Alps node architecture

Alps is operated by Swiss National Supercomputing Centre in Switzerland. This profile covers the documented compute inventory. Specifications are listed below; anything the operator has not published is itemised in the documented-fields list.

Published benchmark (TOP500 · June 2026)

HPL rank
#10
HPL Rmax
434.9 PFlop/s
Theoretical peak Rpeak
574.84 PFlop/s
Ranked core count
2,121,600
Measured power
7124 kW
Energy efficiency
61.05 GFlop/s per watt

Cluster · GH200

Operator / site
Swiss National Supercomputing Centre
Country
Switzerland
Profile scope
Documented compute inventory
Data checked
13 September 2026
Manufacturer
HPE
Ranked core count
2121600
Theoretical peak (PFlop/s)
574.84
Measured power (kW)
7124
Energy efficiency (GFlop/s per watt)
61.0471645143178
Operating system
HPE Cray OS

GH200

Items marked ▧ come from the documented NVIDIA GH200 Grace Hopper Superchip platform specification.

System / node model
HPE Cray EX3000
Compute nodes of this type
2688
CPU model
NVIDIA Grace · 72 cores
CPU sockets per node
4
Cores per socket
72
Physical CPU cores
288
Accelerator model
NVIDIA Hopper (GH200)
Physical accelerators per node
4
Host / unified memory
512 GB LPDDR5X
Accelerator memory
96 GB HBM3 per GPU
Accelerator ↔ accelerator link
NVLink
CPU ↔ accelerator link
NVLink-C2C
Network interfaces per node
4
Interface speed
200 Gb/s injection bandwidth per Grace-Hopper module
Interface-to-accelerator topology
Four Grace-Hopper modules per node with four corresponding network interface cards — a dedicated adapter per accelerator module
Form factor
24 cabinets in 4 rows · 8 chassis per cabinet · 7 blades per chassis · 2 nodes per blade · 112 nodes per cabinet

CPU · Rome

Compute nodes of this type
1024
CPU model
AMD EPYC 7742 · 64 cores
CPU sockets per node
2
Cores per socket
64
Physical CPU cores
128
Host / unified memory
256 / 512 GB DDR (variant unspecified)
Network interfaces per node
2
Interface model
HPE Cray Cassini (Slingshot 11)
Interface speed
200 Gb/s per NIC
Form factor
256 blades · 4 nodes per blade

A100

Compute nodes of this type
144
CPU model
AMD EPYC · 64 cores
CPU sockets per node
1
Cores per socket
64
Physical CPU cores
64
Accelerator model
NVIDIA A100
Physical accelerators per node
4
Host / unified memory
512 GB DDR4
Accelerator memory
80 GB HBM3 per GPU
Network interfaces per node
4
Interface model
HPE Cray Cassini (Slingshot 11)
Interface speed
200 Gb/s per NIC
Interface-to-accelerator topology
Four NICs — a dedicated adapter per accelerator
Form factor
72 blades · 2 nodes per blade

MI300A

Compute nodes of this type
128
CPU model
AMD Zen 4 (MI300A)
Accelerator model
AMD MI300A
Physical accelerators per node
4
Accelerator memory
Unified memory
Accelerator ↔ accelerator link
Infinity Fabric
Form factor
64 blades · 2 nodes per blade

MI250X

Compute nodes of this type
24
CPU model
AMD EPYC · 64 cores
CPU sockets per node
1
Cores per socket
64
Physical CPU cores
64
Accelerator model
AMD MI250X
Physical accelerators per node
4
Host / unified memory
128 GB
Form factor
12 blades · 2 nodes per blade

Scope and uncertainty

Host memory for the A100 partition was corrected to the 512 GB DDR4 that CSCS documents. CSCS marks the MI300A and MI250X node specifications as still to be written, so those partitions carry node counts only. Node specifications follow the documented NVIDIA GH200 Grace Hopper Superchip platform (Superchip module — nodes per system and per-node superchip counts are operator-specific); operator-specific deviations are not published.

Documented fields: 62 / 107

Every field this catalog checks. Fields listed above have a value; the remainder are shown here. A field marked absent is not publicly available in the sources we could verify, which does not mean the hardware is missing. Fields that cannot apply, such as benchmark results for an unranked system, are excluded from the total.

Cluster — 6 / 7

  • Manufacturer
  • Ranked core count
  • Theoretical peak (PFlop/s)
  • Measured power (kW)
  • Energy efficiency (GFlop/s per watt)
  • Operating system
  • Network topology

GH200 — 16 / 22

  • System / node model
  • Compute nodes of this type
  • CPU model
  • CPU sockets per node
  • Cores per socket
  • Physical CPU cores
  • CPU note
  • Accelerator model
  • Physical accelerators per node
  • Host / unified memory
  • Accelerator memory
  • Accelerator ↔ accelerator link
  • On-node fabric
  • CPU ↔ accelerator link
  • Network interfaces per node
  • Interface model
  • Interface speed
  • Interface-to-accelerator topology
  • Aggregate scale-out per node
  • GPUDirect RDMA
  • Management / storage networking
  • Local storage
  • NVMe data disks
  • Max remote data disks
  • Uncached disk IOPS
  • Uncached disk throughput (MBps)
  • Form factor

CPU · Rome — 10 / 15

  • System / node model
  • Compute nodes of this type
  • CPU model
  • CPU sockets per node
  • Cores per socket
  • Physical CPU cores
  • CPU note
  • Accelerator model
  • Physical accelerators per node
  • Host / unified memory
  • Accelerator memory
  • Accelerator ↔ accelerator link
  • On-node fabric
  • CPU ↔ accelerator link
  • Network interfaces per node
  • Interface model
  • Interface speed
  • Interface-to-accelerator topology
  • Aggregate scale-out per node
  • GPUDirect RDMA
  • Management / storage networking
  • Local storage
  • NVMe data disks
  • Max remote data disks
  • Uncached disk IOPS
  • Uncached disk throughput (MBps)
  • Form factor

A100 — 14 / 22

  • System / node model
  • Compute nodes of this type
  • CPU model
  • CPU sockets per node
  • Cores per socket
  • Physical CPU cores
  • CPU note
  • Accelerator model
  • Physical accelerators per node
  • Host / unified memory
  • Accelerator memory
  • Accelerator ↔ accelerator link
  • On-node fabric
  • CPU ↔ accelerator link
  • Network interfaces per node
  • Interface model
  • Interface speed
  • Interface-to-accelerator topology
  • Aggregate scale-out per node
  • GPUDirect RDMA
  • Management / storage networking
  • Local storage
  • NVMe data disks
  • Max remote data disks
  • Uncached disk IOPS
  • Uncached disk throughput (MBps)
  • Form factor

MI300A — 7 / 19

  • System / node model
  • Compute nodes of this type
  • CPU model
  • CPU sockets per node
  • Cores per socket
  • Physical CPU cores
  • CPU note
  • Accelerator model
  • Physical accelerators per node
  • Host / unified memory
  • Accelerator memory
  • Accelerator ↔ accelerator link
  • On-node fabric
  • CPU ↔ accelerator link
  • Network interfaces per node
  • Interface model
  • Interface speed
  • Interface-to-accelerator topology
  • Aggregate scale-out per node
  • GPUDirect RDMA
  • Management / storage networking
  • Local storage
  • NVMe data disks
  • Max remote data disks
  • Uncached disk IOPS
  • Uncached disk throughput (MBps)
  • Form factor

MI250X — 9 / 22

  • System / node model
  • Compute nodes of this type
  • CPU model
  • CPU sockets per node
  • Cores per socket
  • Physical CPU cores
  • CPU note
  • Accelerator model
  • Physical accelerators per node
  • Host / unified memory
  • Accelerator memory
  • Accelerator ↔ accelerator link
  • On-node fabric
  • CPU ↔ accelerator link
  • Network interfaces per node
  • Interface model
  • Interface speed
  • Interface-to-accelerator topology
  • Aggregate scale-out per node
  • GPUDirect RDMA
  • Management / storage networking
  • Local storage
  • NVMe data disks
  • Max remote data disks
  • Uncached disk IOPS
  • Uncached disk throughput (MBps)
  • Form factor

Sources

Browse all cluster architecture pages →

Data & sources